I have a pandas data frame
df = pd.DataFrame({'id':[1,2,3,4],
'attr1':[1,1,0,0],
'attr2':[0,1,1,0],
'attr3':[1,1,1,0],
'attr4':[1,1,1,1]})
I want to convert it to
Basically create a new variable which will contain previous dataframe columns if its value is 1
Use:
df1 = df.filter(like='attr')
df = df.drop(df1.columns, axis=1)
df['var'] = df1.dot(df1.columns + ' ').str.rstrip()
print (df)
id var
0 1 attr1 attr3 attr4
1 2 attr1 attr2 attr3 attr4
2 3 attr2 attr3 attr4
3 4 attr4
Explanation:
filter - only attribute columnsdropDataFrame.dotrstripAlternative solution:
cols = df.columns[df.columns.str.startswith('attr')]
df = df.drop(cols, axis=1).assign(var=df[cols].dot(cols + ' ').str.rstrip())
print (df)
id var
0 1 attr1 attr3 attr4
1 2 attr1 attr2 attr3 attr4
2 3 attr2 attr3 attr4
3 4 attr4
For revert back use str.get_dummies:
df1 = df.join(df.pop('var').str.get_dummies(' '))
print (df1)
id attr1 attr2 attr3 attr4
0 1 1 0 1 1
1 2 1 1 1 1
2 3 0 1 1 1
3 4 0 0 0 1
